80 likes | 169 Views
Probability of admission – further data sources and issues. Colin Cryer IPRU, Univ of Otago, NZ Presented at the ICE on Injury Statistics Meeting, 7-8 September 2006, Washington DC. Other data sources. Issue: Coding frame.
E N D
Probability of admission – further data sources and issues Colin Cryer IPRU, Univ of Otago, NZ Presented at the ICE on Injury Statistics Meeting, 7-8 September 2006, Washington DC
Other data sources Colin Cryer, IPRU, Univ of Otago, NZ
Issue: Coding frame • Several different coding frames used for diagnosis: eg. ICD-9, ICD-9-CM, ICD-10 • Options • Restrict study to just one of these (eg. ICD-10) • Include all • If all, is ICD-9 to ICD-10 translation good enough for comparison? • To resolve Colin Cryer, IPRU, Univ of Otago, NZ
Issue: Specificity of codes • What specificity of ICD diagnosis code should be used to estimate probability of admission? • Issue • Too specific many diagnosis codes with too few cases • Less specific diagnoses with high Pr(Adm) may be hidden in heterogeneous groups. Colin Cryer, IPRU, Univ of Otago, NZ
Issue: Accuracy of coding • Correspondence of ED and hospital discharge diagnosis codes (Source: Farchi, 7th World Conference) • 62.2% of linked cases, diagnosis concordant* • Higher concordance in older people, and less urgent care • Over 80% concordance for: • Fracture (upper and lower extremities, head and neck, spine, torso), lower extremity crushes/ amputation. *At the Barell matrix cell level Colin Cryer, IPRU, Univ of Otago, NZ
Preliminary resultsProbability of Admission – Barell matrixLazio, Italy – Farchi S, Rossi PG, Camilloni L, Chini F (personal correspondence) Source: EIS 2000 - home and road Colin Cryer, IPRU, Univ of Otago, NZ
Probability of Admission – Barell matrix – highLazio, Italy – Farchi S, Rossi PG, Camilloni L, Chini F (personal correspondence) Source: EIS 2000 – home and road Colin Cryer, IPRU, Univ of Otago, NZ
Probability of Admission – Barell matrix – single cellLazio, Italy – Farchi S, Rossi PG, Camilloni L, Chini F (personal correspondence) ICD-9-CM codes with less than 50 ED attendances were excluded Source: EIS 2000 – home and road Colin Cryer, IPRU, Univ of Otago, NZ